Class SequencingQcMetricsServiceImpl

java.lang.Object
ubic.gemma.core.analysis.preprocess.qc.SequencingQcMetricsServiceImpl
All Implemented Interfaces:
SequencingQcMetricsService
@Service public class SequencingQcMetricsServiceImpl extends Object implements SequencingQcMetricsService
Streams report_general_stats_data out of the MultiQC report and resolves its row keys to bioAssays.

Two things about that file drive the implementation:

  • The whole report is around 5.5 MB, of which the general-stats block is about 1% — the rest is report_plot_data. It is read with the streaming parser and every other top-level field is skipped, so the plot data is never materialized.
  • Row keys are not uniformly sample accessions. Measured over 80 reports on the production metadata volume: 3055 rows keyed by the bare accession, 1244 by an accession with a run or mate suffix (GSM5029427_1, GSM5029427_SRR13191146_2), and 1541 by an SRA run accession alone, which Gemma does not record and so cannot be joined. The split falls along module lines — STAR and RSEM key by sample (uniquely_mapped_percent was sample-level in 79 of those 80 reports), FastQC keys by FASTQ file (percent_duplicates in 5 of 80).
Author:
gembro